Blender Git Loki

Fix #32165: context_set_value didn't make an undo push when changing scene settings

This lead to such issues as changing mesh selection mode wasn't doing undo push
when it need to.

Debug option for guarded allocation: store name of original datablock
when using MEM_dupallocN. This helps figuring out issues with non-freed
dup_alloc blocks,

Simply enable DEBUG_MEMDUPLINAME in mallocn.c file.

svn merge ^/trunk/blender -r49175:49177

add prints for texture loading (we're running into texture limit a lot so its handy to see whats actually loading from the blend files)

svn merge ^/trunk/blender -r49167:49175

fix/workaround for glitch with node editing active texture bug - where a node texture in a group could get `stuck` and the buttons UI wouldnt update to other active nodes.

Files saved with this error will still give problems, toggling group edit will fix.

Correction to previous commit -- make sure float buffer is not being changed
while color management transformations are running.

Fix for RNA struct registration: the bpy_class_validate function would only check the immediate functions/properties of the pointer struct type, but not potential base structs. Now it first validates the base struct recursively before the actual properties of the registered class.

Does not have any effect for current registerable types (Operator, Menu, Panel, etc.), since none of those actually have a base struct, but will be required for future types with an actual hierarchy (custom nodes).

Color management: fix memory corruption when using border rendering

Image's get render result function used to modify ImBuf directly,
without making needed invalidates to color management cache when
it's needed.

warn when getting a context member fails because of type mismatch

generalize mask poll functions and sequencer mask code.

Hopefully Blender compiles again with MSVC

Seems to be a conflict between different areas defining
round() function.

Merging r49149 through r49167 from trunk into soc-2011-tomato

Fix corrupted frames producing by fog glare node

Seems to be simple non-initialized buffer used in math, but additional
check would be welcome here.

At least now result doesn't seems to be corrupted and seems to behaving
the same way as non-tile compositor.

no need for renaming actions that are not linked anywhere
talked with Joshua Leung (aligorith) and he agreed on that

use 2d vectors for mask point access.

Tomato Cycles: code reshuffle to make data flow more obvious

Should be no functional changes

mask layer rna api: add exception when removing incorrect masklayer from mask, add mask.layers.clear()

Tomato Cycles: deadlock with save buffers enabled should be resolved now

BGE: Bringing over the dynamic lamp properties fixes from Cucumber (thanks to Daniel Stokes). This means the following KX_LightObject properties now have support when using GLSL materials (in addition to those already supported):
* distance
* lin_attenuation
* quad_attenuation
* spotsize
* spotblend